BGP NextHop field

wuh · ‎11-16-2009

hi,

I was reading CCIE R&S by Wendoll Odom about BGP Nexthop, see the diagram. It first gives a problem in this scenario, "R4's route to 30.0.0.0/8 through R2 lists R1 IP(1.1.1.1)....Unfortunately, R4 doesn't have a route for 1.1.1.1 on R1, so that route cannot be consider best by BGP."

I think R3 SHOULD HAVE a route to 1.1.1.1. why not?

thanks,

Han

Giuseppe Larosa · ‎11-16-2009

Hello Han,

if IP subnet 1.1.1.0/30 is not advertised in iBGP or in an IGP R3 doesn't know about it.

R2 would pass the eBGP route with BGP next-hop = 1.1.1.1 unchanged.

possible solutions are:

advertising ip subnet 1.1.1.0/30 in iBGP with a network command on R2;

advertising ip subnet 1.1.1.0/30 in IGP with a network command on R2;

use of next-hop-self on R2 on session towards R3

Hope to help

Giuseppe

wuh · ‎11-16-2009

Giuseppe

thanks first.

My understanding is that:

as long as R1 advertizes 1.1.1.0/30 to R2. R2 would pass it to all routers in its AS, right?

So, I'd think your condition wouldnt exist. "if IP subnet 1.1.1.0/30 is not advertised in iBGP or in an IGP R3 doesn't know about it. "

the subnet should always be advertised. unless R1 intetinally not to do it by disabling "network" command.

Correct me if I am wrong.

thanks,

Han

Giuseppe Larosa · ‎11-16-2009

Hello Han,

the question is arised for all other prefixes that R1 can advertise to R2 on the eBGP session.

R1 would advertise ip subnet 1.1.1.0/30 with a BGP next-hop of 1.1.1.1, that cannot be installed in other routers in same AS of R2 because BGP next-hop is unknown.

it is a sort of dead lock: prefix 1.1.1.0/30 should be known to accept BGP next-hop 1.1.1.1.

R1 can eventually advertise 1.1.1.0/30 as you noted but BGP next-hop attribute will be 1.1.1.1.

This is not accepted as valid by R2 iBGP peers.

things are different if it is R2 to advertise 1.1.1.0/30.

R2 is not allowed to modify the BGP next-hop over the iBGP sessions unless next-hop-self is used.

I can tell you that this is one of the first problems I had when I started to work on BGP more then 10 years ago.

This is a real problem and not a theorical issue.

Hope to help

Giuseppe